translit.sh

   1 #!/bin/sh
   2
   3 # assumes utf8 locale..
   4 # remove nonascii from the output of pdftotext -layout standard.pdf
   5
   6 sed '
   7 s/\f/(newpage)/g
   8 # utf8 fixes
   9 s/ﬁ/fi/g
  10 s/ﬂ/fl/g
  11 s/§/!S/g
  12 s/©/(C)/g
  13 s/—/--/g
  14 s/−/-/g
  15 s/∗/*/g
  16 s/ˆ/^/g
  17 s/〈/</g
  18 s/〉/>/g
  19 s/⎡/[^/g
  20 s/⎤/^]/g
  21 s/⎣/[_/g
  22 s/⎦/_]/g
  23 s/⎢/[ /g
  24 s/⎥/ ]/g
  25 s/⎧/{/g
  26 s/⎨/{/g
  27 s/⎩/{/g
  28 s/±/(+-)/g
  29 s/≤/<=/g
  30 s/≥/>=/g
  31 s/≠/!=/g
  32 s/Σ/(Sum)/g
  33 s/√/(sqrt)/g
  34 s/π/pi/g
  35 s/∞/(inf)/g
  36 s/ƒ/fl./g
  37 s/∫/(integral)/g
  38 s/Γ/(Gamma)/g
  39 s/×/x/g
  40 s/•/o/g
  41 s/⎯/-/g
  42 s/↑/(uparrow)/g
  43 s/↓/(downarrow)/g
  44 s/↔/<->/g
  45 s/→/->/g
  46 s/‘/'\''/g
  47 s/’/'\''/g
  48 # pdftotext layout fixes
  49 s/_ _/__/g
  50 # floats are sometimes broken
  51 s/\([0-9]\)\. \([0-9]\)/\1.\2/g
  52 ' | LC_ALL=C tr -c '\n-~' '?' | awk '
  53 BEGIN {
  54         getline
  55         last=$0
  56         side=0
  57 }
  58 /^$/ {
  59         nl=nl "\n"
  60         next
  61 }
  62 # TODO: shift page numbers
  63 #function inc(x) {
  64 #       if (x ~ /[0-9]/)
  65 #               return x+1
  66 #       if (sub(/viii$/,"ix",x) ||
  67 #           sub(/iii$/,"iv",x) ||
  68 #           sub(/iv$/,"v",x) ||
  69 #           sub(/ix$/,"x",x))
  70 #               return x
  71 #       return x "i"
  72 #}
  73 /^\(newpage\)/ {
  74         n=split(last,a)
  75         if(side)
  76                 p=a[1]
  77         else
  78                 p=a[n]
  79         side=!side
  80 #       if (p !~ /[0-9]/ && $0 ~ /INTERNATIONAL STANDARD/)
  81 #               p=0
  82 #       print "\n[page " inc(p) "]"
  83         print "\n[page " p "]"
  84         getline
  85         getline
  86         last=$0
  87         next
  88 }
  89 {
  90         print last
  91         last=nl $0
  92         nl=""
  93 }'